Heat loads are the primary input parameters of the aircraft thermal management system, and the real-time prediction of each heat load is significant to achieve the adaptive dynamic control of the thermal management system. In this paper, the real-time inlet and outlet temperature and cooling flow rates of the heat exchanger are constructed in a time-series sequence and used as inputs. The Long-short Term Memory (LSTM) network is applied for the real time prediction of aircraft heat load. The prediction results show that the algorithm can effectively achieve accurate real-time prediction of heat load with a strong practicality.